[PROPOSED EVALUATIONS]
Document extraction evaluation protocol
Field-level correctness, source locations, uncertainty, and reviewer verification.
Proposed evaluation only. No public dataset, completed benchmark run, ranked models, or measured results are available for these tracks.
Proposed evaluation
Use cases with deliberately wrong extractions, correct arithmetic on wrongly mapped inputs, unsupported explanations, and attempts to deliver pending findings. Measure whether reviewers can locate evidence and whether release controls block incomplete states. Track false blocks separately from missed blocks. Human agreement is not assumed to be perfect ground truth. No experimental results or error-elimination claim is made.
Before reporting results
Publish the task definition, data permissions, source coverage, reference-label process, metrics, denominators, and limitations. Separate calculation correctness from professional judgment and disclose unresolved disagreement.